Research On Machine Vision Technology For Measurement Of Shafts' Radial Run-out Error

Machine vision technology uses computer,industrial camera,lens and other auxiliary equipment to simulate visual function,which can implement dimensional measurement,location measurement,shape recognition,color discrimination,defection inspection,OCR/OCV,bar code and QR code recognition.In recently years,it has been obtained more and more attention using this technology to measure the mechanical component's geometrical error because of its flexible non-contact style,high processing efficiency and high-degree integration.Mechanical elements require geometric tolerance for the items that have a significant impact on the element's functional performance.The choice of geometric tolerance is related to the element's machinability directly as well as indirectly reflects its quality and useful life-span.Run-out tolerance is a comprehensive index in the geometric tolerance,which is believed a quality's evaluation for the rolling parts generally.According to the research on the measurement about run-out error at home and abroad,this thesis will put forward a new method to measure the run-out error using the machine vision technology.The thesis starts from practical industrial background and introduces the choice of the main vision measurement equipment and the equipment includes type of cameras,lens,light source which fits into the model presented in this article.Then,the two type of lens distortion models and polynomial camera calibration algorithm on the assumption of pinhole model are described in detail.To correct lens distortion better and improve calibration precision,the initial value of homography matrix which is regarded as an important parameter is studied based on the two-step calibration method.Meanwhile,the direction of optimization plane is studied and the thesis builds the optimization method based on world-coordinate plane.Lastly,the experiment is performed to prove the validity of algorithm.To improve measurement quality of run-out error,the detected algorithm for structured light's center point and the calibration method for light plane are studied.And in order to significantly reduce the calculations burdens and speed up the processing time,the concept of normalized correlation coefficient and image gaussian pyramid are introduced in the plane calibration method,so the improved steg algorithm is put forward.The normal vector calculation based on principal component analysis is also studied and parabolic curve fitting method is implemented in the direction of normal vector avoiding the complexed 2D gaussian convolution calculation further.At last,the experiment for line plane calibration is conducted.It is proved the method in this article is robust and precise by the result of calculation of the size of block gauge.Finally,the run-out error measurement model and reference axis space equation's measurement model are built as well as the method of radial circular run-out error and the whole run-out error are analyzed.And the test bench is set up,the difference between the measurement results from this article and from the dial gauges shows that algorithm put forward in this article is relatively precise.The research has a certain theoretical and engineering value in non-contact measurement on shape and position error for mechanical elements.